Anti-GD2-IRDye800CW as a targeted probe for fluorescence-guided surgery in neuroblastoma

Neuroblastoma resection represents a major challenge in pediatric surgery, because of the high risk of complications. Fluorescence-guided surgery (FGS) could lower this risk by facilitating discrimination of tumor from normal tissue and is gaining momentum in adult oncology. Here, we provide the first molecular-targeted fluorescent agent for FGS in pediatric oncology, by developing and preclinically evaluating a GD2-specific tracer consisting of the immunotherapeutic antibody dinutuximab-beta, recently approved for neuroblastoma treatment, conjugated to near-infrared (NIR) fluorescent dye IRDye800CW. We demonstrated specific binding of anti-GD2-IRDye800CW to human neuroblastoma cells in vitro and in vivo using xenograft mouse models. Furthermore, we defined an optimal dose of 1 nmol, an imaging time window of 4 days after administration and show that neoadjuvant treatment with anti-GD2 immunotherapy does not interfere with fluorescence imaging. Importantly, as we observed universal, yet heterogeneous expression of GD2 on neuroblastoma tissue of a wide range of patients, we implemented a xenograft model of patient-derived neuroblastoma organoids with differential GD2 expression and show that even low GD2 expressing tumors still provide an adequate real-time fluorescence signal. Hence, the imaging advancement presented in this study offers an opportunity for improving surgery and potentially survival of a broad group of children with neuroblastoma.


Anti-GD2-IRDye800CW specifically labels KCNR cells in vitro. Specific binding of anti-GD2-
IRDye800CW was evaluated on the widely used patient-derived NB cell line; SMS-KCNR (KCNR) 16 using an established evaluation pipeline 17 that includes flow cytometry in vitro and fluorescence molecular imaging in vivo (Fig. 1a). Specific binding of anti-GD2-IRDye800CW to NB cells was observed by flow cytometry, with > 95% of cells staining positive, while a negative control colorectal cancer cell line; HT-29, showed no staining, similar to unstained cells (Fig. 1b). In addition, control anti-CD52-IRDye800CW, specific for CD52 (CAMPATH-1) present on the surface of mature lymphocytes, did not label KCNR cells (Fig. 1c). Overall, this demonstrates specific binding of anti-GD2-IRDye800CW to GD2 expressing KCNR cells in vitro.
Effective KCNR-derived tumor detection in vivo and identification of optimal dose and imaging time window. We next addressed the in vivo potential of our probe in a subcutaneous xenograft mouse model. Fluorescence images were generated at multiple days after intravenous administration of 3 different doses of anti-GD2-IRDye800CW. Tumors were visualised with a clinical imaging device that is commonly used for FGS in patients and a preclinical system used for in vivo small animal imaging for fluorescence quantification ( Fig. 1d and Supplementary Fig. S1 online). Fluorescence quantification showed a higher tumor-to-background ratio (TBR) for subcutaneous tumors of mice receiving a dose of 1 nmol and 0.3 nmol anti-GD2-IRDye800CW compared to 3 nmol (Fig. 1e), with a mean fluorescence intensity (MFI) of the tumor significantly higher for 1 nmol dose compared to 0.3 nmol (Fig. 1f). Based on the TBR and MFI curves, we defined an optimal time window for imaging 4 days after administration of 1 nmol anti-GD2-IRDye800CW, with a TBR of 5.2 (SEM ± 1.3) and MFI of 0.28 (SEM ± 0.1), comparable to preclinical TBR and MFI values previously reported for FGS agents successfully used in adult oncology [17][18][19] . Mice receiving the control antibody anti-CD52-IRDye800CW had a significantly lower TBR and MFI, compared to the 1 nmol dose (Fig. 1e, f, Supplementary Fig. S1 online), further demonstrating specific binding of anti-GD2-IRDye800CW to GD2 within tumors. No difference in fluorescence signal was found between tumors of different size ( Supplementary Fig. S2 online and Supplementary Table S1 online).
Orthotopic tumor engraftment demonstrates feasibility for fluorescence-guided surgery. To investigate our tracer in a more clinical setting, we implemented an orthotopic model with KCNR cells transplanted in the adrenal gland, the most common location for NB 20 . Following the optimized conditions determined in the subcutaneous model, mice were intravenously injected with 1 nmol anti-GD2-IRDye800CW and the tumors were resected 4 days post injection guided by the Quest camera ( Fig. 2a-d, Supplementary Video S1 online). In this surgical set-up, we defined a TBR of 6.1 (SEM ± 2.2), similar to the subcutaneous model (Fig. 2e) and detected no remaining fluorescent tissue after surgery (Supplementary Video S1 online). After harvesting the tumors, histology confirmed that NB cells on hematoxylin and eosin (H&E) staining overlapped with fluorescence of anti-GD2 ( Fig. 2f-h), and, importantly, healthy adrenal gland from non-transplanted control mice only showed background levels of anti-GD2-IRDye800CW fluorescence ( Fig. 2i-k). By quantifying the biodistribution, fluorescence in non-tumor tissue was only seen in the femur, as well as in the liver at day 1, which importantly was lower compared to tumor tissue and further diminished by day 4 (Fig. 3a, b), confirming the mostly hepatic clearance of anti-GD2-IRDye800CW. Overall, these data show that anti-GD2-IRDye800CW is a suitable probe for fluorescence tumor detection in vivo.  -GD2 chimeric monoclonal antibody was conjugated to IRDye800CW (left panel) and evaluated in vitro on the NB KCNR cell line by flow cytometry (middle panel). In vivo validation was performed in NB cell line derived xenograft mouse models using the Pearl Trilogy Small Animal imaging system and Quest Spectrum imaging system (right panel). (b) Representative histogram (left panel) and accumulative data (right panel) of anti-GD2-IRDye800CW staining by flow cytometry on KCNR and HT-29 cells, compared to unstained cells. c) Representative histogram (left panel) and accumulative data (right panel) of anti-GD2-IRDye800CW staining on KCNR cells compared to CD52-IRDye-800CW staining. (b, c) n = 3 independent experimental repeats. Graphs depict mean + SEM, ****p < 0.0001. (d) Representative images using the surgical imaging device of mice bearing subcutaneous human KCNR-derived tumors, acquired 1 day (left panel) and 4 days (right panel) after administration of 3 ascending doses of anti-GD2-IRDye800CW. (e) TBR for 7 consecutive days of mice receiving different doses of anti-GD2-IRDye800CW or 1 nmol anti-CD52-IRDye800W as a negative control. Mean ± SEM. ****p < 0.0001; ***p = 0.0006 and **p = 0.0018 for comparison of 1, 0.5, and 3 nmol dose anti-GD2-IRDye800CW, respectively, to 1 nmol dose anti-CD52-IRDye800W. www.nature.com/scientificreports/ Upfront dinutuximab-beta immunotherapy does not interfere with tumor detection. Currently, clinical trials are being initiated with neoadjuvant anti-GD2 immunotherapy prior to surgery in high-risk NB 21 . Therefore, we investigated whether this neoadjuvant immunotherapy would not interfere with the use of anti-GD2-IRDye800CW for FGS. In vitro we showed that 24 h after incubation with a human anti-GD2-FITC antibody (Fig. 4a), KCNR cells could effectively be stained a second time with a human anti-GD2 coupled to a different fluorophore; Phycoerythrin (PE) (Fig. 4b, c). This indicates that GD2 is not lost from the cell surface after antibody binding and that upfront anti-GD2 treatment should not interfere with membranal GD2 fluorescent imaging. To confirm this in vivo, mice with orthotopic induced tumors were pre-treated with a clinically derived dose of 1 nmol anti-GD2 for 2 cycles with 4 days in between, before receiving 1 nmol anti-GD2-IRDye800CW. In this model, the fluorescence signal in mice pre-treated with anti-GD2 dinutuximab-beta was sufficient for www.nature.com/scientificreports/  www.nature.com/scientificreports/ intraoperative fluorescence imaging and tumor resection ( Fig. 4d-g) and TBRs (7.0 SEM ± 3.4) were comparable to those without pre-treatment (6.1 SEM ± 2.2) (Fig. 4h). This demonstrates that the use of neoadjuvant anti-GD2 immunotherapy is unlikely to diminish the fluorescence signal and that anti-GD2-IRDye800CW can be used for FGS in high-risk NB patients regardless of their order of treatment.

Tissue microarray of different pathological tumor stages reveals consistent, yet heterogeneous GD2 expression in high-risk NB patients. To further investigate the extent of patients that can
potentially benefit from anti-GD2-IRDye800CW guided surgery, we defined the expression of GD2 on a tissue microarray (TMA) consisting of tumor samples from 28 high-risk NB patients treated with chemotherapy (Supplementary Fig. S3 online). This demonstrated consistent expression of GD2 across multiple pathological tumor stages; neuroblastoma (Fig. 5a), ganglioneuroblastoma (Fig. 5b) and ganglioneuroma (Fig. 5c) and after chemotherapy, in line with previous literature 13,22 . Importantly, no signal was detected in control peripheral nerve and lymph node tissue (Fig. 5d). However, even when belonging to the same tumor category, heterogeneity between individual patients could be observed ( Supplementary Fig. S3 online) and in each subtype we could identify samples with high, intermediate, or low GD2 expression (Fig. 5a-c).
Patient-derived organoid lines demonstrate adequate tumor detection across differential GD2 expression levels. In the last decade, organoid technology has become a valuable in vitro tool to study human cancer in a patient-specific manner 23 . Since the above results show that GD2 is not uniformly highly expressed in NB patients, we made use of three patient-derived neuroblastoma organoid lines; TIC772 24 , NB67 and NB39 (Supplementary Table S2 and unpublished data) that reflect the high, intermediate and low GD2 expression levels observed in patients, as shown by confocal imaging (Fig. 6a) and flow cytometry (Fig. 6b). Upon subcutaneous transplantation in vivo, tumors derived from each line showed detectable real time fluorescence with differential MFI as observed in vitro (Fig. 6c, d) but, importantly, no difference in TBR compared to KCNR-derived tumors (Fig. 6e). This indicates that even in patients with low GD2 expressing tumors, fluorescence obtained with anti-GD2-IRDye800CW will be sufficient for FGS.

Discussion
By showing the feasibility of intraoperative fluorescence imaging of NB with anti-GD2-IRDye800CW, our preclinical study presents the first suitable molecular-targeted candidate for FGS to guide tumor resection in children. Although FGS with tumor surface antigen specific probes is increasingly implemented in adult oncology 11,25 , its application in pediatric oncology is still scarce. To the best of our knowledge, we here for the first time show that xenograft tumors derived from pediatric tumor cell lines or patient-derived organoids can be detected effectively in vivo using a real time intraoperative imaging system and thereby provide the first step towards closing this translational gap. Importantly, by using a clinical approved antibody combined with a NIR fluorescent probe, we are on a fast track for getting this tumor-specific tracer in a first in-child clinical trial. Before testing efficacy in Phase II/III clinical trials, Phase I clinical trials will be required to determine the safety of anti-GD2-IRDye800CW. However, the immunotherapeutic antibody dinutuximab-beta, one component of our probe, has already been FDA approved. Based on our optimal dose of 1 nmol (0.15 mg) in mice, we expect an effective dose of 0.9 mg/kg (22 mg/m 2 ) in children 26 , close to the dose of 20 mg/m 2 that has been reported to have minimal toxicity in children 27 . In addition, the second part of our probe, the near-infrared dye 800CW has shown to be non-immunogenic in previous preclinical studies 28 and has been safely administrated in multiple clinical trials 29,30 . One important limitation of our preclinical evaluation is that we cannot evaluate binding to GD2 potentially expressed on healthy tissue, because dinutuximab-beta is not cross-reactive with mouse GD2. However, other clinical studies have shown that GD2 expression is restricted to neurons, skin melanocytes, and peripheral sensory nerve fibers 31 with expected intensity signals not-detectable compared to NB. In line with this, we did not obtain positive GD2 staining on control peripheral nerve and lymph node tissue in our TMA. Although safety levels of the conjugated probe still need to be carefully evaluated, these results are highly promising towards the outcome of such evaluations.
Timely progress into clinical trials might benefit children suffering from high-risk NB in multiple ways. Due to the localization of most NB tumors, their encasement in important vasculature and the surgeon's challenge to discern tumor from healthy tissue, resection is almost never complete 4,32 . While standard of care, this also complicates understanding the long-term patient benefit of tumor resection. Indeed, although some studies report that gross total resection improves overall survival 7,33 or event-free survival 34 , others claim no obvious survival benefit 35 . Bias in determining the extent to which tumor resection was complete might contribute to this discrepancy. This is now based on the subjective impression of the surgeon, which is known to have a poor correlation with the results of post-operative imaging 34 . Introduction of FGS for NB resection, will provide an additional modality to quantify tumor cells or tissue remaining, based on fluorescence signal. The extent of resection can thereby be more accurately determined, which opens up new possibilities to reliably assess the effect of surgery on overall survival. In addition, a more accurate quantification of remaining tumor tissue, will help guide decision making on post-surgery course of treatment, and potential benefit of dinutuximab-beta immunotherapy in particular. Most importantly, by providing a precise surgery tool to resect tumoral tissue with higher confidence, chances of relapse due to remaining tumor might decrease. At the same time, damage to surrounding healthy organs and vasculature can be prevented, thereby lowering the risk of surgical complications that is still high in NB patients undergoing surgery 5 .
Considering the overexpression of GD2 on neuroblastoma cells 13   www.nature.com/scientificreports/ even for low expressing organoid lines. Finally, anti-GD2 is increasingly used for immunotherapy in high-risk NB 36 . Currently, after surgery, but ongoing trials will assess the advantage of neoadjuvant treatment 37,38 . Therefore, we also confirmed that anti-GD2-IRDye800CW can still detect NB tissue after upfront anti-GD2 treatment,  www.nature.com/scientificreports/ further validating the wide range of patients that could potentially benefit from anti-GD2-IRDye800CW guided surgery.
In conclusion, we here present a first pediatric cancer specific tracer with potential for the vast majority of high-risk NB patients to guide tumor resection with greater accuracy, thereby lowering the risk of surgical complications and reducing the incidence of relapse. In addition, we envision the comprehensive preclinical evaluation pipeline presented, using both patient derived cell line and organoid xenograft models and encompassing multiple imaging technologies, to be highly applicable for the development of other targeted probes.

Methods
Antibody conjugation. Chimeric monoclonal antibody Dinutuximab-beta (Qarziba, USA) was conjugated to the NIR fluorophore IRDye800CW, (LI-COR Biosciences, Nebraska, USA), as previously described 19 . A degree of labelling (DoL) between 1.0 and 1.5 was considered successful. As a control, the antibody alemtuzumab directed against CD52 present on the surface of mature lymphocytes, was also conjugated to IRDye800CW. Flow cytometry. KCNR and HT-29 were grown to 90% confluency and detached with TrypleLE. Organoids were processed into single cells using 200 ul accutase. Cells were adjusted to 0.5 × 10 6 viable cells per tube in FACS buffer and incubated with 200 μl phosphate-buffered saline (PBS) and 2 μl anti-GD2-IRDye800CW, or anti-CD52-800CW, as a negative control. Organoids were incubated with 2 μl anti-GD2-FITC (mouse 14g2a, Biolegend). After incubation, cells were washed three times in ice-cold PBS and resuspended in 500 μl PBS containing propidium iodine (PI) to stain dead cells. Samples were acquired on a LSRII flow cytometer (BD Biosciences, Singapore) and analysis was performed using FlowJo software (TreeStar, Ashland, Oregon, United States, version 10.6.2).

3D confocal imaging. KCNR cells were transferred to a 96 well sensoplate microplate (Greiner BIO-ONE)
24 h prior to imaging, allowing the cells to form 3D spheroids, due to the low adherence conditions, and incubated with anti-GD2-FITC (mouse 14g2a, Biolegend, 1/200) overnight at 4 °C. The following day, the cells were washed 3 times with medium before incubation with anti-GD2-PE (mouse 14g2a, Biolegend, 1/200) for 15 min on ice.
Patient-derived organoids were incubated directly after culture for 30 min with anti-GD2-FITC on ice before imaging. Imaging was performed on a confocal microscope using a 25X 0.8 NA objective (SP8 Leica microscope, LSM880 Zeiss microscope). 3D rendering was performed using Imaris (Bitplane). Generation of neuroblastoma xenograft models. Mice. Six-week-old athymic nude female mice (CD1-Foxn1 nu , Charles River Laboratories) were used for xenografting of KCNR cells and NSG-mice (bred in house) for organoid xenograft models.

Ethics.
Subcutaneous xenograft models. On average, 1.0 × 10 6 KCNR cells or 1.0 × 10 6 single cells from patient-derived organoids were injected subcutaneously at 2-4 dorsal sites in 50 μl 50% medium/50% BME. Throughout the injection of tumor cells and imaging procedures, animals were anesthetized with 2.5% isoflurane for induction of anesthesia and 2% isoflurane for maintenance with a flow of 0.5 l/min. www.nature.com/scientificreports/ week. When tumors were approximately 8 × 8 mm, mice were injected with anti-GD2-IRDye800CW and imaging performed. In experiments investigating the effect of neoadjuvant dinutuximab-beta treatment, mice were pretreated with 1 nmol in 50 µl PBS 3 weeks after engraftment, followed by a second dose 4 days later. 8 days after the first dose, 1 nmol of anti-GD2-IRDye800CW was administered for subsequent imaging.
In vivo imaging of FGS probes. Mice bearing subcutaneous tumors starting from a size of 8 × 8 mm were intravenously injected in the tail vein with 0.3 nmol, 1 nmol or 3 nmol anti-GD2-IRDye800CW in 50 µl PBS. Fluorescence signal was measured using both the Pearl Trilogy Small Animal imaging system (LI-COR Biosciences, Lincoln, Nebraska, USA) and the Quest Artemis imaging system (Quest Medical Imaging, Middenmeer, The Netherlands). When the mice had multiple tumors, a size of 5 × 5 mm was considered the lower threshold to be included for analysis. Mice bearing organoid-derived subcutaneous tumors, were imaged by the IVIS Spectrum In Vivo Imaging System (Perkin Elmer, Waltham, MA, USA). Control mice were injected with antibody CD52-IRDye800CW (1 nmol). For orthotopic KCNR and patient derived organoid xenograft models, mice were injected with the optimal dose of 1 nmol anti-GD2-IRDye800CW.  22 . Samples were obtained during debulking surgery, formalin fixed, confirmed to represent neuroblastoma tissue by a professor in pediatric oncology pathology and histologically scored as neuroblastoma, ganglioneuroblastoma or ganglioneuroma. Samples were placed on the TMA in duplicate. Control samples of healthy tissue from peripheral nerves and lymphoid tissue were added. The TMA was subsequently stained with anti-GD2 (mouse 14g2a, Biolegend, 1/50).

Histological analysis.
After euthanizing the mice, tumors were surgically removed and fixed in formalin.
Tumors were sectioned and scanned on the Odyssey Clx (LI-COR Biosciences, Lincoln, NE, USA). A solid state laser diode tuned at 785 nm was used for optimal excitation of the fluorophore IRDye800CW and light was collected in the 800 nm channel for evaluation of the fluorescence location. Slides of subsequent sections were stained with haematoxylin-eosin.
Statistical analyses. Statistical analysis was performed using Graphpad Prism software (version 7, Graph-Pad Software Inc, La Jolla, CA, USA). The Area Under the Curve (AUC) was calculated for the different dose groups and comparison of means were performed with the unpaired t-test. All other comparisons of means were performed with the Mann Whitney U test.

Data availability
All data is included in either the main manuscript or Supplementary Information. www.nature.com/scientificreports/